The same technologies that have made our phones smart and computers super are converging on the automobile. For consumers, this will make driving safer, more convenient and less stressful than ever before.
For automakers, this transformation poses significant challenges. Peter Schwarzenbauer, Board member of BMW, recently noted that “the changes in the car industry in the next 10 years will be bigger than in the past 100 years.”
Data gathering and analysis, artificial intelligence, and the pace of innovation are among the areas newly on OEMs’ agenda. Safety must remain core to a carmaker’s approach, but it will also require a robust technology strategy that will ultimately lead to self-driving cars. Thanks to the power and versatile nature of the graphics processing unit (GPU), which is making autonomous vehicles a reality.
From smartphones to smart cars
Rapid advancements in features of smartphones, processing speeds and battery life have come to be expected by consumers on at least a yearly basis. This mobile revolution has raised the expectations of car buyers. To meet consumer demand, auto manufacturers are gearing their efforts to bring cutting-edge technology into vehicles through GPU-powered infotainment and digital dashboard systems.
But these human-machine interface (HMI) applications are just the beginning of what can be achieved with this level of computing power.
Learning to drive with GPUs
It may come as a surprise that the world’s leading automakers and autonomous vehicle researchers are focusing attention on a platform originally developed to play video games. The GPU, invented by Nvidia, is a tiny piece of silicon that punches well above its weight. The highly parallel nature of GPU technology means it’s perfectly suited to processing the type of artificial intelligence required for self-driving cars—deep learning.
The task of driving requires the new computing model of deep learning for conventional programming. Rather than hand-coding software routines with a specific set of instructions to accomplish a particular task, in deep learning the machine is ‘trained’ using large amounts of data and algorithms that give it the ability to learn to perform the task.
To drive itself safely, a car needs to know exactly where it is, recognise the objects around it, and continuously calculate the optimal path. This situational and contextual awareness of a car and its surroundings demands a powerful visual computing system that can merge data from cameras and other sensors, plus navigation sources, while also figuring out the safest path—all in real-time.
Accelerating the race to autonomy
The next generation of vehicles will need to go further, not just collecting but understanding this sensory information using high-performance, energy-efficient processors that can run the search, natural language processing and object recognition algorithms needed for ADAS (advanced driver assistance systems).
Processing these incredible amounts of information requires the power of a supercomputer. But, like those powering smartphones, car-based processors must also be small and operate in an extremely energy-efficient manner. Cramming a desktop-sized computer into a car dashboard is not an option.
From sci-fi to reality
There are good reasons why automakers traditionally run on very long time-lines. Unlike your laptop or smartphone, your car is a life-critical device and needs to undergo incredibly rigorous testing and certification to ensure it’s up to the task. By maintaining these vital standards while calling on the nimble expertise of technology companies, car manufacturers are on the road to proving it’s possible to achieve a happy medium where innovation and safety go hand-in-hand.
For road users, the ability to hand over routine aspects of driving—like sitting in traffic and finding a parking space—to a computer is obviously attractive. But the transformative effects of next-generation car computing will go beyond our day-to-day travel experience. For example, with solid security features in place, the power of the data generated in a single drive could be enormous. Compelling benefits include cheaper insurance, new marketing approaches, reduced accident rates and better road design.
Research already indicates that half of those who buy luxury cars would choose self-driving features. It won’t be long before they become mainstream offerings. In as little as two years, self-driving features will become as ubiquitous as airbags or ABS. And, like these now-standard features, we’ll wonder how we ever managed without them.
The author is managing director, South Asia, Nvidia, the global leader in visual computing